A hardware engineer's guide to DBD actuator design: dielectric stack, electrode geometry, drive waveform, the kilovolt supply, EMC and what fails first.
A surface dielectric barrier discharge actuator has four ingredients: an exposed electrode, a dielectric barrier, a second electrode buried underneath and offset to one side, and an encapsulation layer over the buried side. You can build a working one on a bench in an afternoon out of copper tape and polyimide film. That is what makes the technology deceptive. The part you can sketch in ten seconds is not the part that takes the engineering.
DBD actuator design is the exercise of choosing a dielectric material and thickness, an asymmetric electrode geometry and a high-voltage drive waveform such that the discharge produces useful ionic wind while the barrier that sustains it survives continuous electrical stress. Each choice moves thrust, power, ozone production and service life at once, usually in opposite directions. If you are evaluating plasma actuators for a product, these are the variables you will argue about.
| Variable | Push it this way | You gain | You pay |
|---|---|---|---|
| Dielectric thickness | Thicker | Higher sustainable voltage, higher peak thrust, more ageing margin | Higher drive voltage, larger transformer, thicker package |
| Dielectric permittivity | Higher | More charge per volt, inception at lower voltage | More reactive current, more dissipation in the barrier |
| Exposed electrode edge | Thinner and sharper | Lower inception voltage, more uniform onset | Field concentration, faster electrode erosion |
| Buried electrode width and gap | Wider, gap tuned | Plasma extends further downstream before saturating | Diminishing returns, capacitance the supply must drive |
| Drive frequency | Higher | More discharge cycles per second, more time-averaged thrust | Power rises steeply, more heating, more ozone |
| Drive voltage | Higher | Thrust rises faster than linearly | Power scales roughly with frequency times voltage cubed [3] |
| Duty cycle | Burst modulated | Lower average power and ozone yield | Lower time-averaged thrust |
| Waveform | Nanosecond pulsed, not sinusoidal | Strong thermal and pressure-wave actuation | Almost no ionic wind, so wrong for cooling [3] |
| Encapsulation | Fully potted | Kills discharge on the buried side, excludes contamination | Traps heat, adds capacitance, must not itself age |
The barrier is a capacitor that is also a consumable. Its thickness and permittivity fix the capacitance per unit area, which fixes the charge transferred per cycle, which fixes both thrust and power. Thicker barriers let you apply more voltage before you run out of margin, and the published thrust records reflect that. A four-stage multi-electrode array on 3.175 mm quartz reached roughly 251 mN per metre of span at 45 kV peak to peak and 2 kHz, with a peak wall jet velocity of 7.4 m/s, but drew of order 1,080 W per metre and converted about 0.1 per cent into fluid power [5].
Go the other way and the numbers change completely. A micro-scale actuator on 10 µm of polyimide ran at 5 kV peak to peak and 1 kHz, produced up to 2.0 m/s and about 3 mN per metre of span, and consumed roughly 15 W per metre, at a thrust effectiveness per watt comparable to conventional-scale devices [4]. For electronics thermal management that is the interesting corner of the design space. Nobody is putting a 45 kV supply next to a transceiver cage, and disrupting a boundary layer inside a sealed enclosure does not need 250 mN per metre.
Material choice matters as much as thickness. A durability study of copper electrodes on polyimide, PMMA, quartz glass and alumina found the polyimide actuator failed within 30 minutes through defects that killed the discharge, while copper on alumina settled after a run-in period and held power consumption within about plus or minus 3 per cent over 10 hours [6]. The same work names the attackers: radicals and ozone, temperature, electron bombardment and ultraviolet emission, all acting on the barrier surface at the electrode edge [6]. That is the real lifetime problem, not bulk dielectric strength.
Treat published plasma actuator thrust figures carefully: thrust stand work at NASA Glenn showed electrostatic attraction between actuator and surroundings adding a voltage-dependent artefact to balance readings [1], so figures from different rigs are not automatically comparable.
Asymmetry is what makes the device a pump rather than a lamp: a symmetric arrangement ionises air and produces no net wall jet. Offsetting the buried electrode biases charge transport in one direction, and the resulting body force drags neutral air along the surface. Reviews of surface DBD report typical wall jet velocities of 1 to 10 m/s, with an electrode gap of a millimetre or two a common optimum [3]. Buried electrode width sets how far downstream the plasma extends before the discharge saturates; past that width you are adding capacitance for nothing.
Encapsulation over the buried electrode is not cosmetic either. Without it you get a second, unwanted discharge underneath, wasting power and attacking the dielectric from both faces. With it you have put a thermal blanket over the hottest part of the stack.
The actuator is 200 micrometres thick. What drives it is not. A DBD load is essentially capacitive, and the discharge current is the derivative of the applied voltage, which makes precise control awkward in voltage-fed designs and is why current-fed topologies exist at all [2]. A review of DBD power electronics catalogues the options: resonant inverters with LC or LCL tanks, single-switch class-E stages chosen for component count and voltage gain, and push-pull or bridge current-fed stages where regulated discharge current matters more than cost [2]. Across DBD applications, requirements run from several kV to over 100 kV [2].
Two consequences follow for anyone sizing a plasma actuator power supply. The step-up transformer dominates the volume, because the inter-winding insulation needed for kilovolt isolation sets a floor on its size. That same insulation then fights you electrically: thick, high-permittivity insulation raises parasitic capacitance, which forms a current divider with the load and reduces efficiency, while leakage inductance limits the minimum achievable pulse width [2].
Then there is EMC. A DBD is a train of filamentary microdischarges with sub-microsecond current fronts, on the end of a high-voltage lead that behaves like an antenna. If the product has to pass conducted and radiated emissions testing, the mitigations belong in the first layout: keep the high-voltage run short and shielded, reference the buried electrode to chassis, filter the low-voltage input against common-mode current, and treat creepage and clearance as a spacing rule. Retrofitting shielding onto a working prototype is expensive and rarely enough.
Electromechanical conversion efficiency for surface DBD is fractions of a per cent [5], so this is not a way to move air in bulk. If the problem needs tens of metres per second, a fan is the correct answer and no amount of electrode optimisation changes that. Nanosecond-pulsed drive produces pressure waves rather than ionic wind and will not cool anything [3]. Cutting duty cycle lowers average power and ozone yield, but takes time-averaged thrust with it. And the supply can occupy more volume than the actuator, so an honest comparison against a piezoelectric fan is a comparison of complete assemblies, not of moving elements.
The mechanism has been understood for two decades. What has not been solved generically is longevity: keeping a thin stack stable under continuous electrical stress, at a voltage low enough that the supply stays small, without the barrier degrading at the exposed electrode edge. That is where our research effort concentrates, and it is materials selection, edge definition, encapsulation and drive shaping rather than anything exotic in the physics. The durability data shows why it is worth the effort: two actuators with identical geometry and different barrier materials can differ by orders of magnitude in useful life [6].
For the physics rather than the engineering, start with how DBD plasma actuators work and the wider case for ionic wind as solid-state cooling; ozone and reliability are treated in ionic wind reliability and ozone explained. If you are assessing a specific thermal problem, talk to us with the geometry and the boundary conditions rather than a target airflow figure.
What drive voltage does a DBD actuator need? It depends almost entirely on the barrier. Published micro-scale actuators on 10 µm polyimide operate around 5 kV peak to peak at 1 kHz [4], while millimetre-thick quartz devices used for aerodynamic thrust run up to 45 kV [5]. Thinner barriers mean lower voltage, a smaller supply and less margin against ageing.
Does a thicker dielectric always give more plasma actuator thrust? It raises the achievable maximum, because you can apply more voltage before breakdown, but it does not improve efficiency. Discharge power scales roughly with frequency times voltage cubed [3], so the high-thrust operating points are also the high-power ones.
Can I drive it with a nanosecond pulse generator instead of an AC supply? You can, but you will get a different actuator. Nanosecond-pulsed surface DBD works by depositing heat and launching pressure waves rather than by momentum transfer to neutral air [3], so it is used for aerodynamic flow control, not for generating a steady wall jet.
How large is the plasma actuator power supply relative to the actuator? Larger, usually. The step-up transformer sets the floor on volume because inter-winding insulation is required for kilovolt isolation, and that insulation also adds parasitic capacitance that reduces conversion efficiency [2].
What fails first in a DBD actuator? The dielectric surface near the exposed electrode edge, attacked by radicals, ozone, temperature, electron bombardment and ultraviolet emission [6]. In one comparative study a copper on polyimide actuator failed within 30 minutes, while copper on alumina remained stable over 10 hours of operation [6].